Acupuncture fMRI Findings Across the Alzheimer’s Continuum: What a Critical Review Reveals

A critical review of functional MRI studies has mapped how acupuncture-related brain network findings are distributed across the clinical stages of the Alzheimer’s disease continuum, from subjective cognitive decline through mild cognitive impairment to Alzheimer’s disease dementia. The review, published in Frontiers in Neurology, examined clinical blood oxygen level-dependent functional MRI studies in which acupuncture or electroacupuncture was used as an intervention or stimulation condition in people with these diagnoses. Its central conclusion is cautious: while the reported findings differ in character from one clinical stage to another, those differences cannot be attributed to clinical stage alone, because the underlying studies differ too much in their acupuncture protocols, control conditions, imaging methods, sample sizes, biomarker confirmation, and follow-up.

The review was conducted by Xue Bai, Yahong Ji, and colleagues at Heilongjiang University of Chinese Medicine, with the search of PubMed, the Web of Science Core Collection, and the China National Knowledge Infrastructure last updated on October 14, 2026. The team included original clinical studies of participants with subjective cognitive decline, mild cognitive impairment including amnestic mild cognitive impairment, or Alzheimer’s disease dementia, provided the studies reported acupuncture-related BOLD-fMRI outcomes involving regional spontaneous activity, functional or effective connectivity, or brain-network organization. Both single-session stimulation studies and longitudinal intervention studies were eligible. Studies involving only healthy participants or animal models, studies without separately reported patient data, and studies reporting only structural or other non-BOLD findings were excluded from the primary synthesis, as were conference abstracts, dissertations, and theses.

Because the evidence base was small and heterogeneous, the authors performed no quantitative pooling. Instead, they synthesized the studies narratively by clinical stage, taking into account study design, sample size, acupuncture and control procedures, imaging methods, follow-up, cognitive outcomes, biomarker status, brain-cognition associations, and possible cohort overlap. No formal risk-of-bias or certainty-of-evidence tool was applied; methodological limitations were considered as part of the narrative. Importantly, when multiple reports appeared to arise from the same or overlapping participant cohorts, they were linked and were not treated as independent participant-level evidence, a decision that materially affects how much the literature can actually support.

In subjective cognitive decline, the evidence remains limited and rests on three reports. A small 12-week longitudinal controlled study of 16 participants, with eight receiving acupuncture and eight serving as a no-treatment control, reported decreased regional homogeneity in the prefrontal cortex compared with the control group. A larger sham-controlled randomized trial, with 36 participants per arm, used multi-point acupuncture with electroacupuncture at GV20 and GV24 twice weekly for 12 weeks and found increased hippocampal functional connectivity involving the posterior cingulate cortex and precuneus, the insula, and temporal regions compared with nonpenetrating sham acupuncture. A secondary whole-brain network analysis of the same trial reported significant group-by-time effects on functional connectivity within the sensorimotor network and between the sensorimotor and cingulo-opercular networks, together with changes in parietal nodal efficiency and betweenness. The trial also reported improvement in a composite cognitive score, and the secondary analysis reported improvement in auditory verbal learning test performance and associations between network changes and memory performance. Neither report included Alzheimer’s disease biomarker confirmation.

The mild cognitive impairment literature is the largest body of evidence in the review and spans both acute single-session studies and longitudinal treatment trials. Early work centered on stimulation at the KI3 acupoint in small cohorts of roughly 12 patients and 12 healthy controls, with several reports drawing on the same or substantially overlapping samples. Compared with superficial needling, deep needling was associated with stronger post-stimulus functional connectivity involving the hippocampus, thalamus, and fusiform gyrus, more extensive directed effective connectivity centered on the dorsolateral prefrontal cortex and hippocampus, and modulation of abnormal small-world network properties with increased nodal centrality in regions including the hippocampus, postcentral cortex, and anterior cingulate cortex. A separate six-patient cohort examined GV26 stimulation, with one report describing altered connectivity during needle retention and another describing changes after needle withdrawal. A related ReHo analysis found regional changes after acupuncture, but repeated-measures analysis showed no significant acupuncture main effect or group-by-acupuncture interaction after multiple-comparison correction, underscoring that some of the early findings were not robust.

Later studies in amnestic mild cognitive impairment shifted toward assessing brain function over a course of treatment, with interventions lasting from four weeks to six months. A randomized study of multi-point acupuncture versus sham acupoints over four weeks reported increased connectivity among cognition-related regions including the insula, dorsolateral prefrontal cortex, hippocampus, thalamus, inferior parietal lobule, and anterior cingulate cortex, with the insula, dorsolateral prefrontal cortex, and hippocampus emerging as prominent hubs. A six-month study of manual acupuncture at bilateral LI4 and LR3 found increased right hippocampal connectivity with temporal regions, and lower baseline right hippocampus-to-inferior temporal connectivity was associated with greater subsequent memory change. A sham-controlled eight-week trial reported increased left dorsolateral prefrontal regional homogeneity and altered connectivity, with ReHo change correlating with MoCA change. Other studies reported DLPFC-to-ACC connectivity and effective connectivity changes associated with ADAS-Cog score change, limbic-system connectivity decreases, locus coeruleus-seeded connectivity changes with baseline connectivity predicting later improvement in immediate recall, hippocampal-subregion connectivity changes, and regional ALFF and ReHo changes in a comorbid insomnia subgroup.

In Alzheimer’s disease dementia, most studies used a single session of LR3 and LI4 stimulation and focused on the immediate brain response. These acute studies reported changes in default mode network activity and connectivity involving the posterior cingulate cortex, precuneus, inferior parietal lobule, and middle temporal gyrus; increased hippocampal connectivity including enhanced left hippocampus-to-right middle frontal gyrus connectivity; bidirectional ALFF changes with increased hippocampus-to-precentral gyrus connectivity; and frontoparietal, sensorimotor, and visual network changes. A reanalysis of one dataset using the retrosplenial cortex as a seed found decreased retrosplenial connectivity with the thalamus, posterior cingulate cortex, and other regions in patients but increased connectivity in healthy controls, alongside specifically increased retrosplenial connectivity with the hippocampus and parahippocampal gyrus in the patient group. More recent acute analyses from a separate cohort reported hippocampal-subregion and nucleus basalis of Meynert seed-based connectivity changes. Two longitudinal cohorts extended the picture: a 12-week pilot of acupuncture plus donepezil reported precuneus and temporal fALFF and connectivity differences that did not survive FDR or FWE correction, although temporal fALFF change within the combined-treatment group was associated with ADAS-Cog change; and a six-week randomized study reported stronger left hippocampal connectivity and frontoparietal effective-connectivity changes with acupuncture plus donepezil versus donepezil alone, together with greater MoCA-B improvement in the acupuncture group, though no sham-acupuncture control was used in either cohort.

Across the stages, the review identified a recurring involvement of the hippocampus and posterior medial memory-related regions, including the posterior cingulate cortex, precuneus, and retrosplenial cortex, which overlap with the default mode network and the medial temporal to posterior medial memory system that is vulnerable in Alzheimer’s disease. In mild cognitive impairment the findings spanned a broader range, including memory- and limbic-related regions, DLPFC-ACC connectivity, and locus coeruleus-related pathways, whereas in dementia the findings were more often centered on the hippocampus and posterior medial memory regions. The authors are explicit that these distributions should not be read as stage-specific effects. The mild cognitive impairment evidence includes both single-session KI3 studies and longitudinal treatment studies using several fMRI approaches, while the earlier dementia studies largely used acute LR3/LI4 stimulation and included reports with reused or potentially overlapping samples. The two longitudinal dementia cohorts both examined acupuncture added to donepezil and neither included a sham-acupuncture control, and functional connectivity itself changes across mild cognitive impairment and dementia in the broader imaging literature, further limiting stage-based interpretation.

The review also highlights how the reported changes are not uniformly directional, with both increases and decreases in brain activity and connectivity observed across studies, and it cautions that increases or decreases in activity or connectivity do not by themselves indicate functional improvement or decline, since some changes may reflect compensatory processes. BOLD signal depends on cerebral blood flow and neurovascular coupling, both of which can be altered as Alzheimer’s disease progresses, so acupuncture-related BOLD changes may reflect both neural and vascular factors. Most studies grouped participants by clinical diagnosis with little confirmation of amyloid-beta or tau pathology, meaning the same clinical category may include individuals with different underlying disease biology. The authors call for larger longitudinal studies with biomarker confirmation where possible, appropriate control conditions, complementary imaging modalities such as perfusion imaging, prespecified imaging hypotheses, and cognitive follow-up, to determine whether the observed differences relate to clinical stage and whether the imaging changes persist alongside cognitive change.

Subject of Research: Acupuncture-related functional MRI brain network findings across the clinical stages of the Alzheimer's disease continuum

Article Title: Acupuncture-related brain network findings across clinical stages of the Alzheimer’s disease continuum: a critical review of functional MRI evidence

Article References: Bai, X., Ji, Y., Ruan, H., Zhang, W., Ma, Y., Jiang, X., Liu, X., & Wang, F. (2026). Acupuncture-related brain network findings across clinical stages of the Alzheimer’s disease continuum: a critical review of functional MRI evidence. Frontiers in Neurology, 17, Article 1977763. https://doi.org/10.3389/fneur.2026.1977763

Image Credits: AI Generated

DOI: 10.3389/fneur.2026.1977763

Keywords: acupuncture, Alzheimer's disease, functional MRI, brain networks, mild cognitive impairment, subjective cognitive decline, functional connectivity, hippocampus, default mode network, locus coeruleus, biomarkers, systematic review